Repository navigation
material effects - #2081
material effects#2081
Conversation
|
Hi @sophiemiddleton,
which require these tests: build. @Mu2e/write, @Mu2e/fnalbuild-users have access to CI actions on main. ⌛ The following tests have been triggered for f00cd50: build (Build queue - API unavailable) |
|
☀️ The build tests passed at f00cd50.
N.B. These results were obtained from a build of this Pull Request at f00cd50 after being merged into the base branch at a9ddd06. For more information, please check the job page here. |
oksuzian
left a comment
There was a problem hiding this comment.
PR Review Summary — #2081
Reviewed at head f00cd50e59f3b7cb9a85a7c10a1713ff85694c1e. First pass.
Decision
- 🟡 comment. The geometry, the materials and the new extrapolation code check out. With the two new options turned off, the output is identical to
mainon all 16508 events of the nightlyreco_03input. But both options are on by default inLHDriftFit, which production on-spill reconstruction uses. With them on, about 1% of the LoopHelix fits are lost and almost every fit changes (finding 1). Please settle that before merging: either show it is an improvement, or turn the defaults off.
Scope understood
- What the PR adds:
KinKalGeomgets the calorimeter front panels as three layers each (carbon fiber, foam, carbon fiber), the cooling pipes as centerlines, and the crystal-case inner rings.- These get new surface IDs 320-323.
KKExtrap::extrapolateCaloMaterialadds them when a fit is extrapolated to the calorimeter (ToCaloMaterial).KKFit::addCaloMaterialadds the ones inside the fit range and refits (AddCaloMaterial).KKShellXing::updateReferencenow searches the neighbouring pieces when a crossing moves.MaterialsList.datagets six materials.
- Both options are set
trueinMu2eKinKal/fcl/prolog.fcl, at:395(LHDRIFTXTRAP) and:505(LHDriftFit).KKDe,KKUe,KKDmuandKKUmuinProduction/JobConfig/reco/prolog.fcl:57-84inherit them. SorecoMC/OnSpill.fcl, which runs in 18 prodtools campaign entries, picks them up.OffSpill(KKCHmu) andExtracted(KKLine) keep the C++ default offalse.- The description mentions neither the production defaults nor the
KKShellXingchange.
Findings
- 🟠 [S1] The new options are on in production on-spill reconstruction. With them, about 1% of the LoopHelix fits are lost and almost every fit changes, and the PR has no validation.
- Evidence: the nightly
reco_03input, all 16508 events, onmain(the nightly builda9ddd063) and on this PR merged onto it. The track counts are:
- Evidence: the nightly
| fit | main |
this PR |
|---|---|---|
KKDe |
16486 | 16362 |
KKUe |
16303 | 16146 |
KKDmu |
16316 | 16109 |
KKUmu |
16043 | 15934 |
- Of the 16252 matched
KKDetracks, 16134 change, and 2514 change their active hit count. - For conversion-like
KKDetracks (95-110 MeV atTT_Front, fit consistency above 1e-3, at least 20 hits; 10769 onmain), the change in momentum atTT_Frontis small: median 0.007 MeV, 99th percentile 0.34 MeV. 34 tracks move by more than 1 MeV, and 1 is lost. - Most of the lost tracks are poor fits: 101 of the 125 lost
KKDetracks have fewer than 20 hits. But good fits are lost too. - Example: event 1430:1065:1269, a 54 MeV
KKDetrack with 27 hits before extension.- On
main:After extending the fit : goodFit = 1, fitcon = 0.9054, nHits = 30, 1 calo-hits. - At this head: the three disk-0 front panel layers are added (at 766.2 to 766.5 ns), then
goodFit = 0, fitcon = 0.0000, and the track is not saved.
- On
- The refit is the cause.
KKTrack::addFitMaterialXingscallsextendwith the full fit configuration (KKTrack.hh:334), which replays the whole extension schedule. In the printout, meta-iterations 13 to 21 follow the original 12. That replay is where hits change state and fits fail. - The replay also costs CPU. The four LoopHelix fits take 112.8 ms per event against 81.5 ms on
main, andKKDealone goes from 20.6 to 31.6 ms. - Control: with
AddCaloMaterialandToCaloMaterialset tofalseon the four fits, everyKalSeedcollection, with all its intersections, is identical tomainin all 16508 events. So all of the change comes from these two options. - Suggested fix, any of these:
- Set both defaults to
falseuntil validated, with momentum resolution and efficiency against MC truth on conversion electrons and on a background sample. - Or add the panel material in the same extension step that adds the
CaloHit, rather than in a second full extension.
- Set both defaults to
- Either way, please say in the description that production output changes.
- ⚪ [S3] The pipe and ring layout is a second copy of the G4 construction.
KinKalGeomMaker::makeCaloPipesreproducescaloBuildFrontPlateinconstructDiskCalorimeter.cc, including its0.9975angle factor, the 2 mm shortening and the- 0.1manifold margin (KinKalGeomMaker.cc:199,:211,:215).- If the G4 layout changes, KinKal goes out of step without any error.
- A description of the pipes kept once, in
DiskCalorimeter, and used by both would avoid that. This does not gate.
- ⚪ [S3] Small items.
- The front panel material names come straight from the geometry file and must exist in
MaterialsList.data.geom_common_trackerVST.txt:373-374sets them to"G4_AIR", which is not there.MatDBInfo::findDetMaterialthen throwsCannot find requested material G4_AIRat the first crossing.- The only user,
Production/Processing/vstdata_epilog.fcl, runs line fits with the options off, so nothing fails today.
- Regrown fits do not get the in-fit material (the TODO at
KKFit.hh:418), while the regrow configurations useLHDRIFTXTRAPand so add it in the extrapolation.RefRegrow.fclis in no campaign. DiskCalorimeterMaker.cc:117givesinnerCFRingMateriala default of"CarbonFiber". G4 reads the same key with no default, and both calorimeter geometry files set it, so the default can go.
- The front panel material names come straight from the geometry file and must exist in
Checked, no action needed
- 🟢 The surface IDs are append-only.
- The new values are 320-323, after
EMC_Disk_1_Back(317), so no existing value moves. SurfaceIdis stored as the enum integer throughEnumToStringSparse, so older files read the same.nSurfaceIdsis used only in thestatic_assert(SurfaceId.cc:95).- EventNtuple stores
sidas an integer, with no calorimeter-specific list.
- The new values are 320-323, after
- 🟢 Every new
TimeRangeis built from a time inside the range.extrapolateCaloMaterial:tstartis clamped into the fit range (KKExtrap.hh:491) and later set from intersection times (:515).extrapolateCaloSupport(:572) does the same.caloSupportXings(CaloMaterialLayers.hh:64,:74) starts fromgood()intersections, andgood()requires the intersection to be in range.KKFit::addCaloMaterialorders its ends (KKFit.hh:440).KKShellXinguses ±1 ns (:90).- The calorimeter material is added after the DS and CRV material passes and before the material-free disk extension (
KKExtrap.hh:151). That is the order #2059 needed.
- 🟢 The materials match G4's own definitions (
ConstructMaterials.cc:1266-1295,:1372-1378):CarbonFiber_resin: C5H5O2 at 1.1 g/cm³.CarbonFiber: 46% fiber at 1.422 g/cm³. G4 computes the same weighted density.AluminumHoneycomb: 0.03 g/cm³.CaloPipeAlandCaloRingAl: G4_Al, ascalorimeter.pipeMaterialand the aluminium rings use.
- 🟢 The layer positions match G4.
- The case front is
frontFaceCenter - crystalCapZLength, because the crystal origin sits one cap length behind the case front (DiskCalorimeterMaker.cc:215). G4 places the front plate directly upstream of the case. - The 3 / 21.75 / 3 mm layers and the pipe plane, one pipe radius into the foam, follow
caloBuildFrontPlate(ZposPipe).
- The case front is
- 🟢 With the options off, the
KKShellXingchange and the new geometry leave every output unchanged (the control in finding 1). - 🟢 No crash. All 16508
reco_03events run at this head with its defaults, and so do 200ceSimRecoevents.
Validation check
- Build: Offline
maina9ddd063merged with this head, with Production94a4d81fand mu2e-trig-config2b90c646, p109. These are the same three commits as the nightly build. Production/Validation/nightly/reco_03.fclon its two dig inputs, read in place, all 16508 events, with everyKalSeedcollection printed per event (PrintModuleverbose 2: fit summary and all intersections). It was run three ways:main, this PR, and this PR with the two options off.Production/Validation/ceSimReco.fcl, 200 events (baseSeed8),mainagainst this PR:- 84 of 85
KKDetracks change. - The change in momentum at
TT_Fronthas a median of 0.012 MeV and a maximum of 1.29 MeV. - No track is lost.
- 84 of 85
- With
KKDeprintout on, over 100 events, front panel material was added about 290 times and inner-ring material 21 times. A second pass was needed 4 times. - Not checked:
- Momentum resolution against MC truth.
- The cooling-pipe wall depth (
wallDepth) against G4, beyond reading the code.
Author follow-ups
- Settle finding 1. Either turn the defaults off until validated, or show resolution and efficiency against MC truth, and avoid the second full extension that loses fits.
- Say in the description that production output changes, and mention the
KKShellXingchange.
- KKTrack::addFitMaterialXings refits with a single meta-iteration at the final temperature and no hit updaters, instead of replaying the full extension schedule, which changed hit states and lost good fits - ToCaloMaterial and AddCaloMaterial default to false in prolog.fcl until validated against MC truth, so production reco is unchanged - drop the innerCFRingMaterial default, matching the G4 construction - add G4_AIR to MaterialsList.data for geom_common_trackerVST.txt Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
oksuzian
left a comment
There was a problem hiding this comment.
PR Review Summary — #2081
Reviewed at head 495acb2f2286059d9fa85bc16a1111990f574e2e. Re-review of f00cd50.
Decision
- 🟢 Approvable. The new commit settles my main finding: both options are now off by default, and with the PR's defaults every
KalSeedcollection is identical tomainon all 16508 events of the nightlyreco_03input. The new refit also loses far fewer fits when the options are turned on. FNALbuild has not built this head yet (mu2e/buildtest: "This test has not been triggered yet"), so I am posting this as a comment and have asked for a build. I will approve once it is green. My local build of this head compiles without warnings.
What changed since f00cd50
- One commit, "Address PR #2081 review: calo material refit and defaults". The branch was not rebased; the merge base is still f00cd50. It touches four files:
KKTrack.hh:336-338:addFitMaterialXingsnow refits with a one-entry schedule at the last temperature and no updaters, instead of replaying the whole extension schedule.Mu2eKinKal/fcl/prolog.fcl:395and:505:ToCaloMaterialandAddCaloMaterialarefalse.DiskCalorimeterMaker.cc:117: theinnerCFRingMaterialdefault is gone.MaterialsList.data:9: aG4_AIRentry, a copy ofair.
Prior findings
- 🟢 [was 🟠 S1] The options were on in production on-spill reconstruction. FIXED in 495acb2, verified.
fhicl-dumpat this head ofValidation/nightly/reco_03.fcland ofJobConfig/recoMC/OnSpill.fcl,OffSpill.fclandExtracted.fclshows both optionsfalseonKKDe,KKUe,KKDmu,KKUmuandRegrowLH. Production94a4d81fand mu2e-trig-config2b90c646set neither option.- I reran the same 16508
reco_03events in the same six chunks, with everyKalSeedcollection printed per event. With the PR's defaults, every collection, including all intersections, is identical tomain. The four LoopHelix fits take 81.8 ms per event, against 81.6 ms onmain.
- 🟢 [was ⚪ S3] Small items. FIXED.
G4_AIRis now inMaterialsList.data, so the VST geometries no longer name a material the KinKal material list lacks. I did not run a VST job.- The
innerCFRingMaterialdefault is gone. This cannot break a working geometry: the same maker already readscalorimeter.FPCoolPipeRadiuswith no default, and the only two files that set it (calorimeter_CsI_v1.txt,calorimeter_CsI_v2.txt) also setinnerCFRingMaterial. The production geometry includescalorimeter_CsI_v2.txt, and the 16508-event run built it. - The Regrow TODO (
KKFit.hh:418) is unchanged. With the options off it has no effect.
- ⚪ [S3, open] The pipe and ring layout is still a second copy of the G4 construction (
KinKalGeomMaker.cc:199,:211,:215). This does not gate.
With the options turned on
These numbers are for whoever validates the options later; they do not gate this PR. On the same 16508 events, with both options true on the four LoopHelix fits:
| fit | main |
f00cd50 | 495acb2 |
|---|---|---|---|
KKDe |
16486 | 16362 | 16478 |
KKUe |
16303 | 16146 | 16294 |
KKDmu |
16316 | 16109 | 16249 |
KKUmu |
16043 | 15934 | 16033 |
- Hit states are kept. 0
KKDeand 0KKUmutracks change their hit count, against 2514 and 4837 at f00cd50. - The CPU cost is small: 85.9 ms per event for the four fits, against 112.8 ms at f00cd50.
- For conversion-like
KKDetracks, the momentum change atTT_Fronthas a median of 0.009 MeV and a 99th percentile of 0.15 MeV (0.34 MeV at f00cd50). 24 move by more than 1 MeV. - 🟡 [S2] Some good fits are still lost, through the parameter-divergence test. 4 conversion-like
KKDetracks and 9 conversion-likeKKDmutracks are lost, among 10KKDeand 76KKDmulosses in total.- Example: event 1430:1652:1372, a 99.4 MeV
KKDetrack with 34 hits. After the extension the fit has consistency 0.957. The material is then added: the three disk-0 front-panel layers at 1261.4 to 1261.5 ns, and two disk-1 inner-ring crossings. The single material iteration ends inFit Status ParamsDiverged Meta-iteration 13 iteration 0 ... prob 0.955355, and the track is not saved. With the PR's defaults it is saved (goodFit = 1, fitcon = 0.9567). - 1430:1612:4983
KKDe(prob 0.853) and 1430:1652:560KKDmu(0.774, after two disk-0 inner-ring crossings) fail the same way. - KinKal sets
ParamsDivergedwhen the end parameters move by more thanDivergenceDeltaParams(1e6, as a chi-squared) against the previous fit. The fit just before the material is added is converged. So a 3 mm carbon layer, or a few mm of aluminium ring, should not move the parameters by that much. The cause may be in how the new crossings are seeded, rather than in the one-iteration schedule. I did not find which. - Since the options are off by default, this does not affect production. It is worth settling before the options are turned on.
- Please also mention in the description that both options are off by default. It still describes them only as added options.
- Example: event 1430:1652:1372, a 99.4 MeV
Checked, no action needed
- 🟢 The new refit keeps hit states by construction. With no updaters,
KKStrawHit::updateWHSandKKStrawHitCluster::updateStateleave the wire-hit states as they were.KKStrawXingkeeps its cached updater settings. - 🟢 The increment adds no
TimeRange. The windows in the calorimeter material code are unchanged from f00cd50, which I checked then, and no run raised an exception. - 🟢 Merge state: this head merges cleanly with current
main(858d5f4d). It still conflicts with Offline#2084 (4afa41b1) inMu2eKinKal/inc/KKExtrap.hh(git merge-tree). Whichever merges second needs a rebase.
Validation check
- Build: Offline
maina9ddd063(the nightly build) merged with this head, with Production94a4d81fand mu2e-trig-config2b90c646, p109. It is an incremental muse build of the workspace from my first pass, with exit status 0 and no warnings. - Jobs:
Production/Validation/nightly/reco_03.fclon its two dig inputs, read in place, all 16508 events in six chunks, everyKalSeedcollection printed per event. One run used the PR's defaults, and one had both options on for the four LoopHelix fits. All 12 chunks exited with status 0. Both runs are compared with themainoutputs from my first pass.- Single-event debug reruns of 1430:1652:1372, 1430:1612:4983, 1430:967:4391 (
KKDe) and 1430:1652:560 (KKDmu). fhicl-dumpof the four production and validation configurations above.
- Not checked:
- FNALbuild at this head.
- Momentum resolution against MC truth.
- A VST job with
G4_AIR. - A Regrow job.
- The cause of the
ParamsDivergedfailures.
Author follow-ups
- Before the options are turned on: find why the material refit fails the parameter-divergence test on converged fits, and show resolution and efficiency against MC truth.
- Mention in the description that both options are off by default.
- Coordinate the merge order with Offline#2084, since the two conflict in
KKExtrap.hh.
|
@FNALbuild run build test |
1 similar comment
|
@FNALbuild run build test |
|
for this: Before the options are turned on: find why the material refit fails the parameter-divergence test on converged fits, and show resolution and efficiency against MC truth I will show some plots in the Wednesday meeting. Should I post them here now? |
|
Adding calorimeter passive material differs from upstream passive material
because the calo hit occurs after the material, meaning it is not a pure
extrapolation. Dealing with that requires a different workflow.
…On Mon, Nov 2, 2026 at 7:37 AM Sophie Middleton ***@***.***> wrote:
*sophiemiddleton* left a comment (Mu2e/Offline#2081)
<#2081 (comment)>
for this: Before the options are turned on: find why the material refit
fails the parameter-divergence test on converged fits, and show resolution
and efficiency against MC truth
I will show some plots in the Wednesday meeting. Should I post them here
now?
—
Reply to this email directly, view it on GitHub
<#2081?email_source=notifications&email_token=ABAH572TSV5TKQXD5VO2W7L5XCS4PA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTMNJYGEZDEOJQGA3KM4TFMFZW63VMORSWC3K7NVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-6581229006>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABAH574RALZFS6MVNPXAEBL5XCS4PAVCNFSNUABFKJSXA33TNF2G64TZHMZDAMRYGA2DMOJSHNEXG43VMU5TMMJQG42DKMRUGU32C5QC>
.
You are receiving this because you are on a team that was mentioned.Message
ID: ***@***.***>
--
David Brown ***@***.***
Office Phone (510) 486-7261
Lawrence Berkeley National Lab
M/S 50R5008 (50-6026C) Berkeley, CA 94720
|
|
⌛ The following tests have been triggered for 495acb2: build (Build queue - API unavailable) |
|
☀️ The build tests passed at 495acb2.
N.B. These results were obtained from a build of this Pull Request at 495acb2 after being merged into the base branch at b227150. For more information, please check the job page here. |
|
I agree Dave, it is different. Should the workflow be discussed offline or on Wednesday? |
oksuzian
left a comment
There was a problem hiding this comment.
PR Review Summary — #2081
Reviewed at head 495acb2f2286059d9fa85bc16a1111990f574e2e. Approval following my re-review at this head.
Decision
- 🟢 approve. FNALbuild build 3541 passed at this head when merged with
mainb2271506: the build, every validation job,POT,trigger,g4surfaceCheck,rootOverlaps,check_cmakeand the whitespace check are green. My re-review at this head (review 5730873493) found the earlier S1 fixed: both options are off by default, and with the defaults everyKalSeedcollection is identical tomainon all 16508reco_03events.
Before the options are turned on
- The open items from the re-review still apply then, not now: the parameter-divergence failures on converged fits, resolution and efficiency against MC truth, and the workflow question Dave raised. Plots in a later PR that turns the options on, or in that PR's description, would be the natural place for them.
Merge order, unchanged
- This PR still conflicts with #2084 in
Mu2eKinKal/inc/KKExtrap.hh. Whichever merges second needs a rebase.
|
We can start the discussion Offline and continue it Wednesday if necessary.
…On Mon, Nov 2, 2026 at 8:51 AM Sophie Middleton ***@***.***> wrote:
*sophiemiddleton* left a comment (Mu2e/Offline#2081)
<#2081 (comment)>
I agree Dave, it is different. Should the workflow be discussed offline or
on Wednesday?
—
Reply to this email directly, view it on GitHub
<#2081?email_source=notifications&email_token=ABAH573SVL27OOEH6ELOZK35XC3RLA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTMNJYGI4DSOJYG4Y2M4TFMFZW63VMORSWC3K7NVSW45DJN5XKKZLWMVXHJLDGN5XXIZLSL5RWY2LDNM#issuecomment-6582899871>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABAH576XPOXW7CSQ4NEQH4L5XC3RLAVCNFSNUABFKJSXA33TNF2G64TZHMZDAMRYGA2DMOJSHNEXG43VMU5TMMJQG42DKMRUGU32C5QC>
.
You are receiving this because you are on a team that was mentioned.Message
ID: ***@***.***>
--
David Brown ***@***.***
Office Phone (510) 486-7261
Lawrence Berkeley National Lab
M/S 50R5008 (50-6026C) Berkeley, CA 94720
|
brownd1978
left a comment
There was a problem hiding this comment.
Overall this looks very good, I just have some minor suggestions. Eventually we may want to run a dedicated schedule, not just the last step. And we might want to add more straws at the same time the calo material is added. Both of those are beyond the scope of this PR.
| matcylxings_.push_back(matxingptr); | ||
| } | ||
|
|
||
| template <class KTRAJ> void KKTrack<KTRAJ>::addFitMaterialXings(KKMATRECXINGCOL const& planexings, KKMATCYLXINGCOL const& cylxings) { |
There was a problem hiding this comment.
I suggest naming this addCaloMaterialXings, to be more consistent with the other materials (IPA, ST, ...). The difference is that you need to extend the fit at the end, but that's OK.
There was a problem hiding this comment.
Yes, I can do that
| matconfig.schedule_ = KinKal::Config::Schedule{KinKal::MetaIterConfig(matconfig.schedule().back().temperature())}; | ||
| this->extend(matconfig,nohits,exings); | ||
| for(auto const& matxing : planexings)matrecxings_.emplace_back(matxing); | ||
| for(auto const& matxing : cylxings)matcylxings_.emplace_back(matxing); |
There was a problem hiding this comment.
I suggest to store these new xings in dedicated containers (calomatrecxings, etc), as these are used in the fit not purely in extrapolation.
There was a problem hiding this comment.
Adding this in now
|
Thanks, I'll look into these changes |
oksuzian
left a comment
There was a problem hiding this comment.
PR Review Summary — #2081
Reviewed at head 149c022858acf2fc949451d32386b8679f97f5ce. Re-review of 495acb2.
Decision
- 🟡 comment now, approve when green. The new commit does what Dave asked and changes no fit result. On all 16508 nightly
reco_03events, everyKalSeedcollection prints line for line the same as at 495acb2, both with the PR's defaults and with both options on. My local build of this head has no warnings. FNALbuild has not built 149c022 (mu2e/buildtest: "This test has not been triggered yet"), so I will approve again once a build at this head is green.
What changed since 495acb2
- One commit, "updated for Daves comments", touching three headers (+33 -13). The branch was not rebased; the merge base is still 495acb2.
KKTrack.hh:addFitMaterialXingsis renamedaddCaloMaterialXings. The calorimeter crossings it adds now go into two new containers,calomatrecxings_andcalomatcylxings_, with their own accessors. The cloning constructor copies them (:96-97).KKFit.hhandKKExtrap.hh: the four "already in the fit" checks read the new containers.sampleFitsaves the new containers whenSaveMaterialXingsis set (KKFit.hh:979-992).
Prior findings
- 🟢 Dave's two inline suggestions, the rename and the dedicated containers. FIXED in 149c022, verified.
- Nothing at this head still calls
addFitMaterialXings. The only writers of the new containers areKKTrack.hh:345-346. The extrapolation still writes its own crossings tomatrecxings_andmatcylxings_(:332,:358). - The narrower checks skip exactly what they skipped before. The fit adds the calorimeter material inside
extendTrack(KKFit.hh:407), before the module calls the extrapolation (LoopHelixFit_module.cc:361then:412;RegrowLoopHelix_module.cc:300then:303).extrapolateCaloMaterialruns once per track (KKExtrap.hh:151) and visits each layer and each ring once. So the old containers never held a front-panel or ring crossing that the new ones lack. - I reran the same 16508 events at this head. The printed
KalSeedcontent, the fit summary and every intersection, matches 495acb2 line for line: about 1.86 million lines with the PR's defaults, and about 1.86 million with both options on for the four LoopHelix fits. The defaults run still matchesmainas well. reco_03does not setSaveMaterialXings, so that run does not exercise the newsampleFitloops. The only fit that sets it,KKCHmuinOffSpillandExtracted, leaves both calorimeter options off, so the new containers are empty there.
- Nothing at this head still calls
- 🟢 [was 🟠 S1] The options were on by default. Still fixed:
prolog.fcl:395and:505are unchanged andfalse. Productionmainhas not moved since94a4d81f, and none of mu2e-trig-config's four newer commits mention these options. - 🟡 [S2, carried over] With both options on, some converged fits are still lost through the parameter-divergence test: 4 conversion-like
KKDeand 9KKDmutracks on the 16508 events. The commit does not change the refit, and the options-on run above reproduces the 495acb2 output exactly. Sophie plans to show plots at Wednesday's meeting. This does not gate while the options are off.- The description still does not say that both options are off by default.
- ⚪ [S3, carried over] The pipe and ring layout is still a second copy of the G4 construction (
KinKalGeomMaker.cc). The commit does not touch it, and it does not gate. - Dave's other points are not requests for this PR. In his approval he called a dedicated refit schedule and adding straws "beyond the scope of this PR". The workflow question, that the calorimeter hit lies after the material, is being discussed on the thread and at the meeting.
New in this increment
- ⚪ [S3] Two small things, neither gating.
SaveMaterialXingsnow also controls the calorimeter crossings added in the fit, but its description still says "filled during extrapolation" (KKFitSettings.hh:108,prolog.fcl:42,KKFit.hh:170).sampleFitnow has seven copies of the same six-line block, one per crossing container (KKFit.hh:956-1006). One small helper taking the container would replace them.
Merge order, unchanged
- This head merges cleanly with
main1d88a811. It still conflicts with Offline#2084 (4afa41b1) inMu2eKinKal/inc/KKExtrap.hh(git merge-tree). Whichever merges second needs a rebase.
Validation check
- Build: Offline nightly
maina9ddd063merged with this head, on the CI buildmuseCIBuild/main/a9ddd063_94a4d81f_2b90c646(Production94a4d81f, mu2e-trig-config2b90c646), p109. This is the same base as my 495acb2 runs.muse buildfinished with exit status 0 and no warnings.mainhas not changedMu2eKinKalsincea9ddd063. - Jobs:
Production/Validation/nightly/reco_03.fclon its two dig inputs, read in place, 16508 events in six chunks, with the same job file as before. One run used the PR's defaults, and one had both options on forKKDe,KKUe,KKDmuandKKUmu. All 12 chunks exited with status 0 and no exceptions. - The comparison drops one stdout line,
KalSegment: Invalid time range, tmin 2.68e16 tmax 0, whichKalSegment::timeRange()prints while the input trigger collectionTrigMprKSFDeis being printed. It appears in some runs and not others on the same input, at 495acb2 as well, so it does not come from this commit. - Config contract: this increment changes no fcl and no fhicl parameter.
Not checked
- FNALbuild at this head.
- A job with
SaveMaterialXingsand the calorimeter options both on, which is the only case the newsampleFitloops change. - A Regrow job, a VST job, and resolution against MC truth, as before.
|
@FNALbuild run build test |
|
⌛ The following tests have been triggered for 149c022: build (Build queue - API unavailable) |
|
☀️ The build tests passed at 149c022.
N.B. These results were obtained from a build of this Pull Request at 149c022 after being merged into the base branch at ffcc1a5. For more information, please check the job page here. |
oksuzian
left a comment
There was a problem hiding this comment.
PR Review Summary — #2081
Reviewed at head 149c022858acf2fc949451d32386b8679f97f5ce. Follow-up to my comment review at the same head.
Decision
- 🟢 approve. FNALbuild has now built this head, and every test passed: the prof build, all twelve validation jobs,
trigger,rootOverlaps,g4surfaceCheck,check_cmakeand the whitespace check. This is the approval my comment review promised once the build was green. The open 🟡 about the fits lost when both calorimeter options are on, and the two ⚪ items, stand and do not block while the options are off by default.
This pull request introduces support for modeling and tracking the calorimeter disk front panel material and embedded cooling pipes in the Mu2e simulation and reconstruction. The changes add new surface IDs for the front panel layers and pipes, implement a detailed geometric and material model for these components, and update the extrapolation and configuration logic to include them in fits and track propagation. This enables more accurate modeling of material effects for tracks passing through the calorimeter front panels.
Calorimeter front panel and cooling pipes modeling:
EMC_Disk_0_FrontPanelandEMC_Disk_1_FrontPaneltoSurfaceIdEnum, updated the surface ID count, and registered their string names. [1] [2]CaloFrontPanelPipesclass to represent the geometry and material of cooling pipes embedded in the calorimeter front panels, including methods to compute the material depth crossed by a track. [1] [2]KinKalGeomandKinKalGeomMakerto construct and store the front panel material layers and cooling pipes for each calorimeter disk, including detailed geometric construction and material assignment. [1] [2] [3] [4] [5] [6] [7]Geometry and configuration enhancements:
Extrapolation and material handling:
CaloMaterialLayers) to access both fixed-thickness front panel layers and position-dependent cooling pipe material for use in extrapolation and fitting.